RetentionJuly 20266 min readBy Siddharth Deepak — Founder

LTV Cohort Analysis From a Single CSV Export

You don't need a data warehouse for cohort LTV. One orders export with customer id, date, and net revenue is enough to see whether your acquisition actually pays back.

The only columns you need

customer_id, order_date, and net revenue (after discounts and refunds). Everything else is garnish. From those three you can build the entire cohort table.

The pivot

cohort_month = first_order_month(customer_id)
month_offset = months_between(order_date, cohort_month)
ltv[cohort][offset] = cumulative Σ(net_revenue) / cohort_size

Rows are acquisition months, columns are months since first purchase, cells are cumulative revenue per customer. Read across a row and you watch a cohort mature; read down a column and you compare cohort quality over time.

The two readings that matter

Payback window — the column where cumulative LTV crosses your blended CAC. If March cohorts pay back in month 2 and September cohorts still haven't by month 5, your acquisition quality dropped and blended ROAS never told you.
Cohort decay — are newer rows systematically lower at the same offset? That's product or audience degradation, visible quarters before it shows in revenue.

Traps

Use net revenue — refunds concentrated in early months flatter no one. Exclude incomplete months from comparisons (a 12-day-old cohort isn't "underperforming"). And segment by first-order product when you can: LTV by entry SKU is where merchandising decisions actually live.

Gross LTV is a vanity curve. Contribution-based LTV (revenue minus variable costs) is the one that funds ad budgets.

Computed, not estimated

ProfitFalcon runs this exact math on your store exports — every number verifiable.

Try ProfitFalcon free